https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65841
--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> --- I confirm the segfault with my clean trees for all variants (4.8 and 4.9 give an ICE). However I don't see them on my patched trees having Mikael's patch for pr65792 (https://gcc.gnu.org/ml/fortran/2015-04/msg00063.html) and Andre's one for pr59678 (https://gcc.gnu.org/ml/fortran/2015-04/msg00061.html). Note that the last test in comment 1 segfault with only the [fist patch] for pr65792 + the one for pr59678. All the test run without segfault with the cumulated patch for pr61831 and pr65792 (https://gcc.gnu.org/bugzilla/attachment.cgi?id=35352) + the one for pr59678. Note also that repeated assignments is checked by gfortran.dg/alloc_comp_assign_10.f90 (pr39879): u%model = u%design u%model = u%design ! The double assignment was the cause of the ICE